Automated speech recognition (ASR) systems attempt to determine the content of speech. Many ASR systems use an acoustic model to predict which sounds are present in audio data, and use a language model to predict which words were spoken.